Procurement Optimization in Motor Insurance

Procurement optimization in motor insurance describes the systematic improvement of purchasing decisions in the claims process. For total economic losses, the core question is: which salvage platform should the accident vehicle be sold through to achieve the highest return?

In Germany, there are several specialized salvage exchanges, each with its own customer base. Each platform has strengths for certain vehicle types. Choosing the right platform for the right vehicle yields measurably higher returns. AI-driven procurement optimization makes this selection data-based — instead of relying on habit or individual contracts.

A machine learning model analyzes historical bid data from all relevant salvage platforms. Based on vehicle characteristics such as class, manufacturer, age, mileage, and replacement value, the model predicts which platform will yield the highest bid for a specific vehicle.

In practice, this means: instead of defaulting to a single platform, each vehicle is assessed individually. When prediction confidence is high, the best platforms are targeted. When uncertain, the full range is used. The result is intelligent routing that minimizes costs and maximizes returns.

The reasons are historical: existing contracts, established processes, lack of comparative data. Many market participants simply have no systematic overview of which platform delivers the best results for which vehicle type.

On top of that: the differences between platforms are not obvious. Only an analysis of hundreds of thousands of bid records reveals the patterns — on certain platforms, higher bids are placed for SUVs, on others for older vehicles, on yet others for specific manufacturers. Without data analysis, these differences remain invisible.

Two core datasets: vehicle appraisal data (vehicle class, manufacturer, replacement value, age, mileage, engine power, damage type) and historical bid data from salvage platforms (bid amount, platform, accepted bid). This data typically already exists within insurers but is rarely analyzed systematically.

The real work lies in integration: appraisal databases and bid databases are often separate systems with different formats. Merging and cleaning these data silos is a prerequisite for any analysis.

The model is accurate enough that, on average across all vehicles, it demonstrably achieves higher returns than any blanket platform assignment. It operates efficiently even with large data volumes and delivers interpretable results.

What matters in practice: the model doesn't just deliver a prediction, but also a confidence score. At high confidence, it routes precisely. At lower confidence, it recommends broader distribution across multiple platforms.

The benchmark is competitors that use only a single salvage platform. Systematically selecting the economically best platform for every vehicle disposal yields higher residual values per vehicle on average.

Across hundreds of thousands of claims per year, this difference adds up to around 50 million euros annually. The calculation is based on real bid data and a comparison between actually achieved returns and a simulation where only the most commonly used single platform would have been used.

The core principle — data-driven optimization of vendor selection — is applicable across industries. Wherever companies regularly choose between multiple service providers, suppliers, or platforms, a similar approach can be implemented.

The prerequisite is a sufficient data foundation of historical transactions. The more decisions are documented, the more precisely a model can identify patterns and derive recommendations. The automotive sector is particularly well-suited because data volumes are large and features are well-structured.

When a vehicle is declared a total economic loss — when repair costs exceed the vehicle's replacement value — the vehicle is auctioned through specialized salvage exchanges. Buyers are typically commercial purchasers, workshops, and recyclers.

The residual values achieved directly impact the insurer's claims costs: the higher the residual value, the lower the compensation to the policyholder, the lower the loss ratio. Platform selection is therefore a business-critical decision — one that has rarely been made systematically until now.